AMD Radeon Cloud China Public Free Model APIs (incl. DeepSeek V4.1 Flash)
China developer portal lists free OpenAI-compatible endpoints for DeepSeek-V4-Flash, V4-Flash-Vision-Exp, V4.1-Flash, Qwen3.8-Flash-Next, Qwen3.8-27B, GLM-5.3-Flash, MiMo-V2.6-Flash, MiniCPM5-2B, plus MinerU2.5-Pro, OmniVoice, Clef, d3, d3-flash.
Last verified · UTC+8
Last scanned · UTC+8
自动核验尚未通过,后续将自动重试。
下次自动复核:2026-11-13 06:00 UTC+8
查看核验原因
部分线索缺少官方证据,已安排自动复核
Offer & quota details
已核验模型用途:Writing & chat、Coding、Vision & OCR。具体免费范围仍以活动规则为准。 用途依据 ↗ 用途依据 ↗ 用途依据 ↗ 用途依据 ↗
- Access method
- Use the OpenAI-compatible API at https://developer.amd.com.cn/radeon/api/v1
Show official excerpt
每个模型一页,对应 https://developer.amd.com.cn/radeon/api/v1 上提供的模型。
- Verification & payment
- Direct call, no payment or card binding (China public API)
Show official excerpt
对应 https://developer.amd.com.cn/radeon/api/v1 上提供的模型。 当前有哪些模型以 GET /v1/models 为准
- Limits & model scope
- All chat models on /v1/chat/completions except OCR/speech/decision; reasoning_effort values vary per model
Show official excerpt
除以下五个各走自己的端点外,其余都在 /v1/chat/completions 上应答
- Usage & data policy
- Presence on the page does not imply the model is live; rely on GET /v1/models
Show official excerpt
这里有页面并不代表该模型当前在线。
- Claim deadline
- No explicit claim deadline published
API integration
Official documentation checks and live API tests are separate. This site has not claimed these offers or tested the APIs for you.
- API Base URL
https://developer.amd.com.cn/radeon/api/v1- Protocol
- See the official integration documentation
- Model IDs / scope
DeepSeek-V4-FlashDeepSeek-V4-Flash-Vision-ExpDeepSeek-V4.1-FlashQwen3.8-Flash-NextQwen3.8-27BGLM-5.3-FlashMiMo-V2.6-FlashMiniCPM5-2BMinerU2.5-ProOmniVoiceClefd3d3-flash- Official website可访问
- Unknown / Not tested
- 账号可注册
- Unknown / Not tested
- China服务器调用已验证
- Unknown / Not tested
- 是否需要信用卡
- Unknown / Not tested
下一步:核对items件与API region
选择符合账号items件的 API,再View model details与配置;Ended活动可以继续查看同类Active活动。
Sources & verification
AMD Radeon Cloud China区 · Monitoring enabled
每个模型一页,对应 https://developer.amd.com.cn/radeon/api/v1 上提供的模型。 当前有哪些模型以 GET /v1/models 为准Read official terms
Your official account determines eligibility and remaining credits. Deadline alerts use explicit claim dates, not inferred dates from credit validity.
Verified changes
2026/10/17 06:00:18
| 变化项目 | 之前 | 核验后 |
|---|---|---|
| 模型范围 | ["DeepSeek-V4.1-Flash","DeepSeek-V4-Flash","DeepSeek-V4-Flash-Vision-Exp","Qwen3.8-Flash-Next","Qwen3.8-27B","MiniCPM5-2B"] | ["DeepSeek-V4-Flash","DeepSeek-V4-Flash-Vision-Exp","DeepSeek-V4.1-Flash","Qwen3.8-Flash-Next","Qwen3.8-27B","GLM-5.3-Flash","MiMo-V2.6-Flash","MiniCPM5-2B","MinerU2.5-Pro","OmniVoice","Clef","d3","d3-flash"] |
| 额度 | 官方免费共享端点,不消耗 GPU Credits,无固定 Token 赠额上限;每日请求量按账号动态配置,超额直接返回 429,不会自动转为付费扣费。 | 未记录 |
| 领取资格 | 中国区独立账号(developer.amd.com.cn/radeon),API Key 需通过平台审核后获取。 | 未记录 |
| 申请条件 | 仅限评测、实验、原型、学习、Demo、Hackathon 与开源开发;明确禁止生产负载,明确禁止代理/网关/商业转售;请求日志滚动保留约 2 天(不用于模型训练)。 | 直接调用,无需付费或绑卡(中国区公开 API) |
| 用完后计费 | 永久零单价共享端点,超额直接报错 429,无静默扣费风险。 | 未记录 |
| 限流与范围 | 典型限流 30 RPM / API Key、120 RPM / IP、8 并发、网关 20 RPM / Account;实际限额以 API 响应为准。 | 除 OCR/语音/决策外都在 /v1/chat/completions;reasoning_effort 取值随模型而异,max_tokens 需按上下文窗口调整 |
